In recent years, with an increase in integration of semiconductor elements and a reduction in size of a semiconductor element, a technique for quickly and accurately inspecting a fine pattern becomes important. From such a background, with sophistication of an imaging device represented by an electron microscope and a video device, an image processing device has been put into practical use which has: a function to recognize a specific pattern included in an image, the image being obtained by imaging a sample with an imaging device, and the location thereof and carry out alignment between the specific pattern and a pattern on previously registered data (hereinafter, this function is referred to as “pattern matching”); and a function to measure the dimensions of a target pattern (hereinafter, this function is referred to as “pattern measurement”).
Patent Literature 1 discloses an invention related to a method of precisely measuring a shape error, called a process fluctuation in a pattern forming step, between a pattern on an image of a sample and a pattern on previously registered data of the sample, wherein if the pattern measurement result is not appropriate, the size and/or location of a cursor are corrected.
Patent Literature 2 discloses a method including the steps of: when a pattern is measured by setting, imaging, and image-processing a target region including an edge of the pattern, calculating a shape difference between a pattern in a SEM image and a pattern in design data; based on this calculated information, specifying an inspection region with a measuring cursor; and managing the location and/or shape thereof as a measurement recipe, wherein if the location of the cursor is offset from the edge, the location of the cursor is shifted in accordance with the location of the edge so as to increase the distance between the edge and the cursor.
Patent Literature 3 discloses a method including the steps of: setting a mask region, which is excluded from a subject for pattern matching, in the image data of a sample; and if previously registered pattern data spans a target region and the mask region in the image data in a scanning stage, setting a processing condition based on the extent of overlapping with at least one of the target region and the mask region.